Portal de Periódicos da CAPES

Crystal Structure of the Intrinsically Flexible Addiction Antidote MazE

2003; Elsevier BV; Volume: 278; Issue: 30 Linguagem: Inglês

10.1074/jbc.m302336200

1083-351X

Remy Loris, Irina Marianovsky, Jurij Lah, Toon Laeremans, Hanna Engelberg–Kulka, Gad Glaser, Serge Muyldermans, Lode Wyns,

Enzyme Structure and Function

A specific camel VHH (variable domain of dromedary heavy chain antibody) fragment was used to crystallize the intrinsically flexible addiction antidote MazE. Only 45% of the polypeptide chain is found ordered in the crystal. The MazE monomer consisting of two β-hairpins connected by a short α-helix has no hydrophobic core on its own and represents only one half of a typical protein domain. A complete domain structure is formed by the association of two chains, creating a hydrophobic core between two four-stranded β-sheets. This hydrophobic core consists exclusively of short aliphatic residues. The folded part of MazE contains a novel DNA binding motif. A model for DNA binding that is consistent with the available biochemical data is presented. A specific camel VHH (variable domain of dromedary heavy chain antibody) fragment was used to crystallize the intrinsically flexible addiction antidote MazE. Only 45% of the polypeptide chain is found ordered in the crystal. The MazE monomer consisting of two β-hairpins connected by a short α-helix has no hydrophobic core on its own and represents only one half of a typical protein domain. A complete domain structure is formed by the association of two chains, creating a hydrophobic core between two four-stranded β-sheets. This hydrophobic core consists exclusively of short aliphatic residues. The folded part of MazE contains a novel DNA binding motif. A model for DNA binding that is consistent with the available biochemical data is presented. In bacteria programmed cell death is known to be mediated through a unique genetic system called "addiction module," which was found to be located in a number of extrachromosomal elements (low copy number plasmids and prophages). (for reviews see Refs. 1Jensen R.B. Gerdes K. Mol. Microbiol. 1995; 17: 205-210Crossref PubMed Scopus (282) Google Scholar, 2Yarmolinsky M.B. Science. 1995; 267: 836-837Crossref PubMed Scopus (294) Google Scholar, 3Couturier M. Bahassi E.M. Van Melderen L. Trends Microbiol. 1998; 6: 269-275Abstract Full Text Full Text PDF PubMed Scopus (116) Google Scholar, 4Engelberg-Kulka H. Glaser G. Annu. Rev. Microbiol. 1999; 53: 43-70Crossref PubMed Scopus (319) Google Scholar). An "addiction module" consists of a pair of genes. The downstream gene encodes a stable toxic protein, whereas the upstream gene encodes a short-lived antidote that antagonizes the toxin. When bacteria loose the plasmid(s) (or other extrachromosomal elements), the cured cells are selectively killed because the unstable antitoxin is degraded faster than is the stable toxin. Thus, the cells are "addicted" to the continued presence of the non-chromosomal genetic element (2Yarmolinsky M.B. Science. 1995; 267: 836-837Crossref PubMed Scopus (294) Google Scholar). Therefore, along with other very precise mechanisms for preventing plasmid loss, such as plasmid partition (5Hiraga S. Annu. Rev. Biochem. 1992; 61: 283-306Crossref PubMed Scopus (145) Google Scholar, 6Williams D.R. Thomas C.M. J. Gen. Microbiol. 1992; 138: 1-16Crossref PubMed Scopus (161) Google Scholar), these addiction modules have a role in maintaining the stability in the host of the extrachromosomal elements on which they are born. Gene pairs, homologous to some of the gene pairs of these extrachromosomal addiction modules have been found in the Escherichia coli chromosome (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar, 8Gotfredsen M. Gerdes K. Mol. Microbiol. 1998; 29: 1065-1076Crossref PubMed Scopus (264) Google Scholar, 9Masuda Y. Miyakawa K. Nishimura Y. Nishimura, Y. Ohtsubo E. J. Bacteriol. 1993; 175: 6850-6856Crossref PubMed Google Scholar, 10Masuda Y. Ohtsubo E. J. Bacteriol. 1994; 176: 5861-5863Crossref PubMed Google Scholar, 11Metzger S. Dror I.B. Aizenman E. Schreiber G. Toone M. Friesen J.D. Cashel M. Glaser G. J. Biol. Chem. 1988; 263: 15699-15704Abstract Full Text PDF PubMed Google Scholar). The mazEF system, located in the E. coli relA operon, was the first described prokaryotic chromosomally born regulable addiction module. It is responsible for programmed cell death in bacteria (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar, 8Gotfredsen M. Gerdes K. Mol. Microbiol. 1998; 29: 1065-1076Crossref PubMed Scopus (264) Google Scholar, 9Masuda Y. Miyakawa K. Nishimura Y. Nishimura, Y. Ohtsubo E. J. Bacteriol. 1993; 175: 6850-6856Crossref PubMed Google Scholar, 10Masuda Y. Ohtsubo E. J. Bacteriol. 1994; 176: 5861-5863Crossref PubMed Google Scholar, 11Metzger S. Dror I.B. Aizenman E. Schreiber G. Toone M. Friesen J.D. Cashel M. Glaser G. J. Biol. Chem. 1988; 263: 15699-15704Abstract Full Text PDF PubMed Google Scholar, 12Engelberg-Kulka H. Reches M. Narasimhan S. Schoulaker-Schwarz R. Klemes Y. Aizenman E. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1998; 95: 15481-15486Crossref PubMed Scopus (94) Google Scholar). The mazEF system has all of the properties that are required for an extrachromosomal "addiction module" (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar, 13Marianovsky I. Aizenman E. Engelberg-Kulka H. Glaser G. J. Biol. Chem. 2001; 276: 5975-5984Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar): (i) MazF is a toxic protein that is antagonized by the anti-toxic protein MazE; (ii) MazF is long-lived, whereas MazE is easily degraded by the protease ClpPA; (iii) MazE and MazF are co-expressed and interact; (iv) MazE is synthesized in excess over MazF; and (v) mazEF is weakly autoregulated by MazE and efficiently autoregulated by the combined action of MazE and MazF. In addition, the mazEF system has several unique properties. Its promoter has an unusual DNA structure called an "alternating palindrome," and it also carries a binding site for the factor for inversion stimulation. In addition, mazEF-mediated cell death can be triggered by several stress conditions that inhibit mazEF expression: (i) high concentrations of guanosine 3′,5′-bispyrophosphate (ppGpp, the product of RelA protein under extreme amino acid starvation) (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar); (ii) antibiotics that are general inhibitors of transcription and/or translation (14Sat B. Hazan R. Fisher T. Khaner H. Glaser G. Engelberg-Kulka H. J. Bacteriol. 2001; 183: 2041-2045Crossref PubMed Scopus (199) Google Scholar); and (iii) the toxin from the phage P1-borne Phd/Doc addiction module (15Hazan R. Sat B Reches M. Engelberg-Kulka H. J. Bacteriol. 2001; 183: 2046-2050Crossref PubMed Scopus (67) Google Scholar). It has therefore been proposed that mazEF plays a role in programmed cell death under stress conditions (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar, 14Sat B. Hazan R. Fisher T. Khaner H. Glaser G. Engelberg-Kulka H. J. Bacteriol. 2001; 183: 2041-2045Crossref PubMed Scopus (199) Google Scholar, 15Hazan R. Sat B Reches M. Engelberg-Kulka H. J. Bacteriol. 2001; 183: 2046-2050Crossref PubMed Scopus (67) Google Scholar, 16Engelberg-Kulka, H., Sat, B., and Hazan, H. (2001) ASM News, in pressGoogle Scholar). MazE and MazF are homologous to PemI (Kis) and PemK (Kid), respectively, the antidote and toxin encoded by the pemIK addiction system of plasmid R100 (identical to the kis/kid system on plasmid R1) (9Masuda Y. Miyakawa K. Nishimura Y. Nishimura, Y. Ohtsubo E. J. Bacteriol. 1993; 175: 6850-6856Crossref PubMed Google Scholar, 17Bravo A. de Torrontegui G. Diaz R. Mol. Gen. Genet. 1987; 210: 101-110Crossref PubMed Scopus (129) Google Scholar). It has been shown that the mazEF and pemIK systems can interact functionally. MazE partially complements temperature-sensitive mutants of PemI that have lost their ability to prevent cell growth inhibition in the presence of PemK (18Santos-Sierra S. Giraldo R. Diaz-Orejas R. FEMS Microbiol. Lett. 1997; 152: 51-56Crossref PubMed Google Scholar). The cellular target of the MazF toxin is still unknown. DnaB was found to be the target of Kid of the kis/kid system of plasmid R1 (19Ruiz-Echevarria M.J. Gimenez-Gallego G. Sabariegos-Jarego R. Diaz-Orejas R. J. Mol. Biol. 1995; 247: 568-577PubMed Google Scholar). The availability of structural data on addiction proteins has been very limited. The structure of the toxin CcdB of the ccdAB system was the first to be reported, in 1999 (20Loris R. Dao-Thi M.-H. Bahassi E.M. Van Melderen L. Poortmans F. Liddington R. Couturier M. Wyns L. J. Mol. Biol. 1999; 285: 1667-1677Crossref PubMed Scopus (126) Google Scholar), and recently the parD-encoded Kid toxin has been shown to have a very similar three-dimensional structure (21Hargreaves D. Santos-Sierra S. Giraldo R. Sabariegos-Jareno R. de la Cueva-Mendez G. Boelens R. Díaz-Orejas R. Rafferty J.B. Structure. 2002; 10: 1425-1433Abstract Full Text Full Text PDF PubMed Scopus (76) Google Scholar), indicating evolutionary relationships between different families of addiction modules. Crystallization of other addiction toxins is hampered by problems of expression. Low thermodynamic stability and very short shelf life have prevented crystallization of the antidotes. In the present work, we have crystallized the chromosomal antidote MazE in complex with a specific camel VHH 1The abbreviation used is: VHH, variable domain of dromedary heavy chain antibody. antibody fragment that was purposely used as a crystallization aid. The resulting crystal structure is the first of any addiction antidote. Expression and Purification of MazE—To obtain pure His tag-fused MazE protein we used the QIAexpressionist kit (Qiagen). The mazE gene was amplified with PCR. The PCR product was digested with BglII and HindIII before its introduction into the BamHI and HindIII sites of the pQE-30 vector. The MazE protein expressed from this construct bears the His tag and enterokinase sites on its N terminus. The pQE-30-mazE plasmid was introduced into E. coli MC4100ΔmazEF-lacIq (relA+) by CaCl2 transformation. Transformants were grown at 37 °C in 1 liter of LB medium supplemented with 100 μm ampicillin and 50 μm kanamycin until A 600 reached 0.6. The mazE gene was induced by the addition of 1 mm isopropyl-1-thio-β-d-galactopyranoside. After a further 3-h incubation cells were harvested, resuspended in 80 ml of lysis buffer (50 mm sodium phosphate, pH 8.0, 300 mm NaCl, 20 mm imidazole, 1 tablet of Complete EDTA-free (Roche Molecular Biochemicals)), sonicated, and centrifuged at 10000 × g for 15 min. The supernatant was loaded on a nickel-nitrilotriacetic acid column, washed with the same buffer, and eluted by an 20–300 mm imidazole gradient. Fractions containing MazE were identified on SDS-PAGE, pooled and dialyzed against double distilled water, and finally lyophilized and stored at –20 °C. The mutants R8A and R16A were constructed by PCR-based site-directed overlap extension mutagenesis according to Ho et al. (22Ho S.N. Hunt H.D. Horton R.M. Pullen J.K. Pease L.R. Gene. 1989; 77: 51-59Crossref PubMed Scopus (6963) Google Scholar) and purified in the same way as the wild-type protein. Expression and Purification of the Anti-MazE VHH Fragment, cAbmaz1—Dromedary immunization and construction of a phage display library were done as described previously (23Lauwereys M. Arbabi Ghahroudi M. Desmyter A. Kinne J. Holzer W. De Genst E. Wyns L. Muyldermans S. EMBO J. 1998; 17: 3512-3520Crossref PubMed Scopus (405) Google Scholar). Phages expressing the VHH repertoire as a gene III fusion protein were rescued from the library using M13K07 helper phages, essentially as described by Arbabi Ghahroudi et al. (24Arbabi Ghahroudi M. Desmyter A. Wyns L. Hamers R. Muyldermans S. FEBS Lett. 1997; 414: 521-526Crossref PubMed Scopus (603) Google Scholar). To identify MazE binders, periplasmic fractions of independent TG1 clones were applied on 2 μg/ml MazE-coated microtiter plates. Specific anti-MazE clones were developed with a mouse anti-Myc tag followed by an anti-mouse alkaline phosphatase conjugate. Binders were reamplified and the PCR fragments were NcoI/BstEII-digested and cloned in pHEN6, a pHEN4-derived expression vector in which the hemagglutinin tag and gene III were replaced by a histidine tag followed by a termination codon (23Lauwereys M. Arbabi Ghahroudi M. Desmyter A. Kinne J. Holzer W. De Genst E. Wyns L. Muyldermans S. EMBO J. 1998; 17: 3512-3520Crossref PubMed Scopus (405) Google Scholar). The recloned binders were expressed in the periplasm of E. coli WK6 and purified as described by Lauwereys et al. (23Lauwereys M. Arbabi Ghahroudi M. Desmyter A. Kinne J. Holzer W. De Genst E. Wyns L. Muyldermans S. EMBO J. 1998; 17: 3512-3520Crossref PubMed Scopus (405) Google Scholar). One single VHH fragment, termed cAbmaz1, with high affinity for MazE, was selected for further experiments. Crystallization and Data Collection—MazE and cAbmaz1 were mixed in a 1:2 ratio, and the complex was purified by gel filtration to remove unbound MazE and possible unreactive MazE fragments. The pure complex subsequently concentrated to a final concentration of 9 mg/ml in phosphate-buffered saline. Good quality crystals were directly obtained in screen condition 18 of the Hampton crystal screen (20% polyethylene glycol 8000, 0.1 m sodium cacodylate, pH 6.5, 0.2 m magnesium acetate). X-ray data were collected on EMBL beamline BW7B of the Deutsches Elektronen Synchrotron (DESY, Hamburg, Germany) to 2.3 Å and subsequently to 1.65 Å on EMBL beamline ID14-3 of the European Synchrotron Radiation Facility (ESRF, Grenoble, France). Crystals were flash-frozen to 100 K directly without the need of additional cryoprotectant. The crystal tried at the ESRF diffracted to 1.65 Å resolution. Half-way through the data collection, technical problems at the beamline resulted in the loss of the crystal. Several other crystals were subsequently tested without success. The Hamburg and Grenoble data sets were subsequently merged to give the data set described in Table I, which was used for structure determination and refinement. These data are thus 99% complete between 15 and 2.3 Å resolution and 58% complete between 2.3 and 1.65 Å. Using such a data set is superior to using only the lower resolution 2.3-Å data set, as it adds a large amount of additional unique data against which the model is refined despite the low completeness in the higher resolution bins. The resulting structure is thus not a true 1.65-Å resolution structure.Table IData collection and refinement statisticsData set IData set 2CombinedData collectionBeamlineBW7B, DESYID14-3, ESRFResolution range (outer shell)15.0-2.3 (2.4-2.3)15.0-1.65 (1.71-1.65)15.0-1.65 (1.71-1.65)(Å)Number of reflections measured50,973 (4,459)45,415 (4,348)96,388 (4,348)Number of unique reflections16,140 (1,570)24,962 (2,633)32,375 (2,633)Mean I/σl12.22 (5.19)6.45 (3.76)11.82 (3.76)R merge0.060 (0.232)0.077 (0.228)0.070 (0.228)Completeness99.3 (98.6)57.9 (61.0)74.1 (61.0)RefinementProtein Data Bank ID code1mvfR crystal0.2177R free0.2491r.m.s.d.aRoot-mean-square deviation. bond lengths0.006 År.m.s.d. bond angles1.36°Ramachandran plot (%)Core91.3Additionally allowed8.0Disallowed0.6a Root-mean-square deviation. Open table in a new tab Structure Determination—The asymmetric unit of the crystals contains a dimer of MazE and two VHH antibody domains. Molecular replacement with AMORE (25Navaza J. Acta Crystallogr Sect. A. 1994; 50: 157-163Crossref Scopus (5038) Google Scholar) using the coordinates of an anti-RNase A VHH fragment (with its three complementarity-determining region (CDR) loops removed) (26Decanniere K. Desmyter A. Lauwereys M. Ghahroudi M.A. Muyldermans S. Wyns L. Struct. Fold. Des. 1999; 7: 361-370Abstract Full Text Full Text PDF Scopus (141) Google Scholar) as a model yielded two clear solutions, resulting in a correlation coefficient of 0.27 and an R-factor of 0.51 using data between 10.0 and 4.0 Å. After rigid body refinement with AMORE followed by positional and B-factor refinement with crystallography NMR software (CNS version 1.0 (27Brunger A.T. Adams P.D. Clore G.M. DeLano W.L. Gros P. Grosse-Kunstleve R.W. Jiang J.S. Kuszewski J. Nilges M. Pannu N.S. Read R.J. Rice L.M. Simonson T. Warren G.L. Acta Crystallogr. Sect. D Biol. Crystallogr. 1998; 54: 905-921Crossref PubMed Scopus (17023) Google Scholar)) (using all data between 22.0 and 1.65 Å), an R-factor of 0.44 and an R-free factor of 0.46 were obtained. Phases from molecular replacement were used as a starting point for the warpNtrace option of ARP (28Perrakis A. Harkiolaki M. Wilson K.S. Lamzin V.S. Acta Crystallogr. Sect. D Biol. Crystallogr. 2001; 57: 1445-1450Crossref PubMed Scopus (462) Google Scholar). After 300 cycles, an almost complete trace of the backbone of the two VHH fragments as well as part the MazE dimer (a hairpin, an isolated β-strand, and two helices) was obtained. Applying the known crystallographic symmetry operator then gave a backbone model containing 60 residues of the MazE dimer. In addition, the wARP map was of excellent quality allowing us to build all side chains of the VHH as well as most of the ordered part of MazE. Subsequently, refinement cycles using crystallography NMR software (initially including a simulated annealing protocol, but at the final stages only positional and B-factor refinement) were alternated with manual rebuilding using TURBO (29Roussel A. Cambillau C. Silicon Graphic Geometry Partners' Directory. Silicon Graphics, Mountain View, CA1989: 71-78Google Scholar) to locate the remaining missing residues as well as ordered water molecules. The final structure has r = 0.218, R-free = 0.249 and has excellent stereochemistry, as shown in Table I. Coordinates and x-ray data for the MazE-cAbmaz1 complex have been deposited at the Protein Data Bank and will be available as entry 1mvf. Structure Analysis—The stereochemical quality of the final crystallographic model was analyzed with PROCHECK (30Laskowski R.A. MacArthur M.W. Moss D.S. Thornton J.M. J. Appl. Crystallogr. 1993; 26: 283-291Crossref Google Scholar). Secondary structure assignments were based upon PROMOTIF (31Hutchinson E.G. Thornton J.M. Protein Sci. 1996; 5: 212-220Crossref PubMed Scopus (1007) Google Scholar). Solvent accessibilities were calculated using the program NACCESS. 2S. J. Hubbard and J. M. Thornton, University College London. Homologues of MazE were searched with BLAST (33Altschul S.F. Gish W. Miller W. Myers E.W. Lipman D.J. J. Mol. Biol. 1990; 215: 403-410Crossref PubMed Scopus (73420) Google Scholar) in the combined Swiss Protein and trEMBL data bases (34Bairoch A. Apweiler R. Nucleic Acids Res. 2000; 28: 45-48Crossref PubMed Scopus (2443) Google Scholar), and the multiple sequence alignment was obtained with ClustalW (35Thompson J.D. Higgins D.G. Gibson T.J. Nucleic Acids Res. 1994; 22: 4673-4680Crossref PubMed Scopus (56652) Google Scholar). The variability at each amino acid position of the multiple alignment was scored using a sum-of-pairs approach based upon the amino acid distance matrix of Miyata et al. (36Miyata T. Miyazawa S. Yashunaga T. J. Mol. Evol. 1979; 12: 219-236Crossref PubMed Scopus (346) Google Scholar) and modified by Armon et al. (37Armon A. Graur D. Ben-Tal N. J. Mol. Biol. 2001; 307: 447-463Crossref PubMed Scopus (396) Google Scholar). A Camel Antibody as Crystallization Aid for a Largely Unstructured Protein—MazE is an 82-amino acid protein (98 amino acids in the used His-tagged construct), but in our crystal structure only 44 residues located in the N-terminal half of the sequence are found ordered. SDS-gel analysis of redissolved crystals nevertheless indicated that in the crystal the protein is intact. Lack of structure of the C-terminal half is thus not a consequence of proteolytic degradation of the sample. This makes MazE a member of the growing class of intrinsically unstructured proteins that become structured upon binding their natural partners. Such proteins are often observed in the areas of cell cycle control and in both transcriptional and translational regulation (38Wright P.E. Dyson H.J. J. Mol. Biol. 1999; 293: 321-333Crossref PubMed Scopus (2387) Google Scholar). The large fraction of unstructured polypeptide explains why this small protein would not crystallize on its own. In complex with the antibody fragment (henceforth called cAbmaz1), the total amount of structured polypeptide (126 amino acids of antibody and 44 amino acids of MazE) rises to 73% compared with 45% of free MazE, thus providing a much better starting point for crystallization. Indeed, the antibody domains should aid in the crystallization of many proteins that contain large portions of unstructured polypeptide, similar to their earlier use in crystallization of membrane proteins (39Hunte C. Michel H. Curr. Opin. Struct. Biol. 2002; 12: 503-508Crossref PubMed Scopus (162) Google Scholar). Fig. 1 shows the packing of the MazE-cAbmaz1 complex in the crystal. The repeating unit is a MazE dimer sandwiched between two cAbmaz1 molecules. All interactions that stabilize the crystal lattice are between cAbmaz1 molecules. The only contacts made by the MazE dimer are those with combining sites of cAbMaz1. A large fraction of unstructured polypeptide in MazE may be of functional importance because it is a common feature of many unrelated addiction antidotes. Preliminary NMR assignments of ParD, the antidote of the parDE system on plasmid RK2/RP4, indicate a folded N-terminal half and a largely unstructured C-terminal half (40Oberer M. Zangger K. Prytulla S. Keller W. Biochem. J. 2002; 361: 41-47Crossref PubMed Scopus (35) Google Scholar). Also for CcdA on plasmid F, thermodynamic unfolding data suggest that the protein may be partially unfolded in vivo (41Dao-Thi M.-H. Messens J. Wyns L. Backmann J. J. Mol. Biol. 2000; 299: 1373-8613Crossref PubMed Scopus (29) Google Scholar). Phd, the antidote of the phd/doc system on phage P1, was found essentially unfolded at 37 °C but contains significant tertiary structure at 4 °C (42Gazit E. Sauer R.T. J. Biol. Chem. 1999; 274: 2652-2657Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar). ParD, CcdA, as well as Phd gain structure upon complexation with operator DNA (41Dao-Thi M.-H. Messens J. Wyns L. Backmann J. J. Mol. Biol. 2000; 299: 1373-8613Crossref PubMed Scopus (29) Google Scholar, 42Gazit E. Sauer R.T. J. Biol. Chem. 1999; 274: 2652-2657Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar, 43Oberer M. Lindner H. Glatter O. Kratky C. Keller W. Biol. Chem. 1999; 380: 1413-1420Crossref PubMed Scopus (33) Google Scholar). The MazE Monomer Corresponds to Only One-half of a Protein Domain—The crystal structure of the MazE monomer is shown in Fig. 2a. It consists of two β-hairpins connected by a short α-helix. This fold is different from the predominantly α-helical secondary structure derived from the NMR data of ParD, the antidote of the parDE system on plasmid RK2/RP4 (40Oberer M. Zangger K. Prytulla S. Keller W. Biochem. J. 2002; 361: 41-47Crossref PubMed Scopus (35) Google Scholar), arguing against a common ancestor for the different families of addiction modules. The MazE monomer does not contain a buried hydrophobic core as such. If the MazE monomer is considered in isolation, all amino acid side chains are solvent-exposed. The protein resembles only half of a protein domain. In the crystal, MazE forms a dimer (Fig. 2b), as it does also in solution (44Lah J. Marianowsky I. Glaser G. Engelberg-Kulka H. Kinne J. Wyns L. Loris R. J. Biol. Chem. 2003; 278: 14101-14111Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar). In this dimer both the N-terminal and the C-terminal hairpins of the two monomers associate to form two four-stranded antiparallel β-sheets. The architecture as judged from the intersheet angle and intersheet separation is quite common for a β-sandwich. The fold itself, however, is unique and does not match any of the proteins found in the protein data bank. Dimerization buries about 1700 Å2 of surface area of each MazE monomer, of which 1360 Å2 (80%) is hydrophobic. This dimer-dependent hydrophobic core consists exclusively of small aliphatic side chains (Val-6, Val-15, Ile-17, Leu-27, Val-33, Leu-42, Ile-44). This hydrophobic core contacts several other hydrophobic residues, which make it extend to the surface of the protein (Pro-13, Pro-18, Leu-21, Leu-37). The only aromatic residue within the ordered part of MazE (Trp-9) is also involved in dimer formation. It is, however, not part of the hydrophobic core and is only partially buried. The MazE Family of Antidotes—The proteins of the maz system show sequence similarity to another chromosomal addiction system, chpB (9Masuda Y. Miyakawa K. Nishimura Y. Nishimura, Y. Ohtsubo E. J. Bacteriol. 1993; 175: 6850-6856Crossref PubMed Google Scholar), to similar systems in the genomes of several bacteria (45Mittenhuber G. J. Mol. Microbiol. Biotechnol. 1999; 1: 295-302PubMed Google Scholar), and to the plasmid R100 (R1)-borne pem (kis/kid) system (17Bravo A. de Torrontegui G. Diaz R. Mol. Gen. Genet. 1987; 210: 101-110Crossref PubMed Scopus (129) Google Scholar, 46Tsuchimoto S. Ohtsubo H. Ohtsubo E. J. Bacteriol. 1988; 170: 1461-1466Crossref PubMed Scopus (121) Google Scholar). Fig. 3 shows an alignment of the sequences of known MazE-like antidotes with the evolutionary conservation of amino acid residues and side chain accessibilities mapped upon the sequence of MazE. Except for the residues that form the hydrophobic core, the best conserved residues (Arg-8, Gly-10, and Ser-12) are found in the N-terminal hairpin. The residues on the solvent-exposed surface of the C-terminal hairpin on the other hand are poorly conserved. They include some of the most variable residues and also accommodate the insertion of a single amino acid in the sequence of Ralstonia solanacearum. The homologous Kis (PemI) protein from plasmid R1 was shown to consist of two functional domains. Deletion of the last 33 amino acids results in a protein that loses its antidote properties but largely retains its regulatory function (47Santos-Sierra S. Pardo-Abarrio C. Giraldo R. Diaz-Orejas R. FEMS Microbiol. Lett. 2002; 206: 115-119Crossref PubMed Google Scholar). Mutations at positions 10, 11, 12, and 18 on the other hand abolish the regulatory activity of Kis but not its potential to antagonize Kid. Thus, the ordered part of MazE most likely is a DNA-binding module, whereas the disordered C-terminal half interacts with MazF. DNA Recognition Domain—The complex of toxin and antidote of all addiction systems tested acts as a repressor for their own synthesis (13Marianovsky I. Aizenman E. Engelberg-Kulka H. Glaser G. J. Biol. Chem. 2001; 276: 5975-5984Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar, 48Tam J.E. Kline B.C. Mol. Gen. Genet. 1989; 219: 26-32Crossref PubMed Scopus (74) Google Scholar, 49Roberts R.C. Spangler C. Helinski D.R. J. Biol. Chem. 1993; 268: 27109-27117Abstract Full Text PDF PubMed Google Scholar, 50Magnuson R. Yarmolinsky M.B. J. Bacteriol. 1998; 180: 6342-6351Crossref PubMed Google Scholar, 51Ruiz-Echevarria M.J. Berzal-Herranz A. Gerdes K. Diaz-Orejas R. Mol. Microbiol. 1991; 5: 2685-2693Crossref PubMed Scopus (71) Google Scholar). The antidotes have a DNA binding activity, which is enhanced in the presence of the toxin (13Marianovsky I. Aizenman E. Engelberg-Kulka H. Glaser G. J. Biol. Chem. 2001; 276: 5975-5984Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar, 42Gazit E. Sauer R.T. J. Biol. Chem. 1999; 274: 2652-2657Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar, 49Roberts R.C. Spangler C. Helinski D.R. J. Biol. Chem. 1993; 268: 27109-27117Abstract Full Text PDF PubMed Google Scholar, 52Dao-Thi M.-H. Charlier D. Loris R. Maes D. Messens J. Wyns L. Backmann J. J. Biol. Chem. 2002; 277: 3733-3742Abstract Full Text Full Text PDF PubMed Scopus (67) Google Scholar, 53Afif H. Allali N. Couturier M. Van Melderen L. Mol. Microbiol. 2002; 41: 73-82Crossref Scopus (120) Google Scholar). The toxins on the other hand have no DNA binding ability on their own. Many different structural motifs are known to mediate DNA binding, but none of them matches the structural features of MazE. The most obvious way for a dimeric protein to interact with a stretch of double-stranded DNA is to align its molecular 2-fold axis with the dyad of the DNA double strand. This means that two potential DNA binding surfaces have to be considered for MazE: the N-terminal and the C-terminal pair of β-hairpins. Simple visual docking showed that the N-terminal but not C-terminal pair of β-hairpins forms a surface that seems to be complementary in shape to the major groove of a piece of B-DNA (Fig. 4a). In addition, Arg-8 and Arg-16 are ideally positioned to interact with a phosphate group of the DNA backbone. This model was tested by site-directed mutagenesis and is backed up by biochemical evidence. When comparing the sequences of the different members of the MazE family of proteins (Fig. 3), the N-terminal hairpins contain a larger number of positively charged residues, including the conserved Arg-8, resulting in a surface with a significant positive electrostatic potential (Fig. 4b). The C-terminal hairpin on the other hand is less conserved and contains many negatively charged residues. We further constructed the mutants R8A and R16A and measured their DNA binding properties with fluorescence and isothermal titration calorimetry (Fig. 5). Although the wild-type protein can clearly be observed to interact with a 50-bp piece of promoter/operator DNA containing an alternating palindrome with a 3:1 MazE:DNA stoichiometry (13Marianovsky I. Aizenman E. Engelberg-Kulka H. Glaser G. J. Biol. Chem. 2001; 276: 5975-5984Abstract Full Text Full Text PDF PubMed Scopus (118) Google Scholar, 44Lah J. Marianowsky I. Glaser G. Engelberg-Kulka H. Kinne J. Wyns L. Loris R. J. Biol. Chem. 2003; 278: 14101-14111Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar), the R16A mutant is essentially inactive (Fig. 5). The R8A mutant displays an intermediate behavior. CD spectra indicate that both mutants remain properly folded. Our model also predicts that bound cAbmaz1 should not interfere with DNA binding. Indeed, the presence of cAbmaz1 has no effect on the DNA binding properties of MazE (Fig. 5). MazF-induced Conformational Changes—While this paper was under review, the structure of a hexameric MazE2MazF4 complex was published (54Kamada K. Hanaoka F. Burley S.K. Mol. Cell. 2003; 11: 875-884Abstract Full Text Full Text PDF PubMed Scopus (258) Google Scholar). In this structure, the complete MazE protein is ordered. The N-terminal domain adopts a structure identical to our VHH complex, showing that the interaction with the antibody does not significantly disturb the structure of this unstable and in solution partially unfolded domain. The C-terminal domain of MazE adopts an extended polypeptide conformation that is embraced and stabilized by MazF. This fully explains the absence of structure for the C-terminal domain in the absence of MazF (as in our VHH complex), the large amount of random coil found in isolated MazE and other antidotes in solution (40Oberer M. Zangger K. Prytulla S. Keller W. Biochem. J. 2002; 361: 41-47Crossref PubMed Scopus (35) Google Scholar, 41Dao-Thi M.-H. Messens J. Wyns L. Backmann J. J. Mol. Biol. 2000; 299: 1373-8613Crossref PubMed Scopus (29) Google Scholar, 42Gazit E. Sauer R.T. J. Biol. Chem. 1999; 274: 2652-2657Abstract Full Text Full Text PDF PubMed Scopus (58) Google Scholar, 43Oberer M. Lindner H. Glatter O. Kratky C. Keller W. Biol. Chem. 1999; 380: 1413-1420Crossref PubMed Scopus (33) Google Scholar, 44Lah J. Marianowsky I. Glaser G. Engelberg-Kulka H. Kinne J. Wyns L. Loris R. J. Biol. Chem. 2003; 278: 14101-14111Abstract Full Text Full Text PDF PubMed Scopus (36) Google Scholar) and the high vulnerability for proteolytic cleavage in the C-terminal halves of all antidotes studied until now (7Aizenman E. Engelberg-Kulka H. Glaser G. Proc. Natl. Acad. Sci. U. S. A. 1996; 93: 6059-6063Crossref PubMed Scopus (507) Google Scholar, 55Van Melderen L. Dao-Thi M.-H. Lecchi P. Gottesman S. Couturier M. Maurizi M.R. J. Biol. Chem. 1996; 271: 27730-27738Abstract Full Text Full Text PDF PubMed Scopus (188) Google Scholar, 56Lenherr H. Yarmolinsky M.B. Proc. Natl. Acad. Sci. U. S. A. 1995; 92: 3274-3277Crossref PubMed Scopus (158) Google Scholar) Conclusion—Programmed cell death in bacteria is typically mediated by so-called addiction modules that encode for a stable toxin and a labile antidote. During normal growth, both proteins are expressed at low levels and form a complex that acts as a repressor for their own synthesis. Induction of the system involves rapid degradation of the antidote by a specific protease. The toxins interfere with strategically important proteins such as gyrase or DnaB that are also relevant targets for antibiotics. Several families of addiction systems are known to reside on low copy number plasmids and phages and in the genomes of E. coli and other bacteria. With the exception of the ccd system on plasmid F, biochemical characterization remains limited in most cases and structural information is available only for the toxin CcdB of the ccd module. Structural analysis of addiction proteins has been hampered by the potential expression levels of the toxins and by the intrinsic flexibility of the antidotes. Here we use a specific camel VHH antibody fragment to crystallize the addiction antidote MazE. Only 45% of the polypeptide chain is found ordered in the crystal, making it a member of the growing class of proteins that contain intrinsically flexible domains. The MazE monomer has no hydrophobic core on its own and represents only one-half of a typical protein domain. A complete domain structure is formed by the association of two chains, creating a hydrophobic core between two four-stranded β-sheets. The folded part of MazE constitutes a new DNA binding motif. A model for DNA binding involving the N-terminal hairpins of both subunits is backed up by additional evidence from biochemical and protein engineering experiments. We acknowledge the use of the synchrotron beamtime at the EMBL beamlines at the DORIS storage ring (Hamburg, Germany) and the European Synchrotron Radiation Facility (Grenoble, France). We thank Joris Messens and Karolien Van Belle for purifying the MazE-VHH complex.